Institutional Ownership and Return Predictability Across Economically Unrelated Stocks

Transcription

1 Cornell University School of Hotel Administration The Scholarly Commons Conference Proceedings, Presentations, and Speeches School of Hotel Administration Collection Institutional Ownership and Return Predictability Across Economically Unrelated Stocks George P. Gao Cornell University, Pamela Moulton Cornell University School of Hotel Administration, David T. Ng Cornell University, Follow this and additional works at: Part of the Finance and Financial Management Commons Recommended Citation Gao, G. P., Moulton, P. C., & Ng, D. T. (2015, July). Institutional ownership and return predictability across economically unrelated stocks [Electronic version]. Retrieved [insert date], from Cornell University, SHA School site: This Article is brought to you for free and open access by the School of Hotel Administration Collection at The Scholarly Commons. It has been accepted for inclusion in Conference Proceedings, Presentations, and Speeches by an authorized administrator of The Scholarly Commons. For more information, please contact

2 Institutional Ownership and Return Predictability Across Economically Unrelated Stocks Abstract We document strong weekly lead-lag return predictability across stocks from different industries with no customer-supplier linkages (economically unrelated stocks). Between 1980 and 2010, the industry-neutral long-short hedge portfolio earns an average of over 19 basis points per week. This return predictability arises exclusively from pairs of stocks in which there are common institutional owners. This predictability is a new phenomenon which does not originate from the slow information diffusion underlying previously documented lead-lag effects, weekly reversals, momentum, nonsynchronous trading, or other known factors. Our findings suggest that institutional portfolio reallocations can induce return predictability among otherwise unrelated stocks. Keywords institutional ownership, return predictability, anomalies, institutional trading Disciplines Finance and Financial Management Comments Required Publisher Statement Copyright held by the authors. Original paper presented at the 23 rd Annual Conference on Financial Economics and Accounting November 16 and 17, 2012, Los Angeles, CA. This article is available at The Scholarly Commons:

3 Institutional Ownership and Return Predictability Across Economically Unrelated Stocks George P. Gao, Pamela C. Moulton, and David T. Ng* July 10, 2015 Abstract We document strong weekly lead-lag return predictability across stocks from different industries with no customer-supplier linkages (economically unrelated stocks). Between 1980 and 2010, the industry-neutral long-short hedge portfolio earns an average of over 19 basis points per week. This return predictability arises exclusively from pairs of stocks in which there are common institutional owners. This predictability is a new phenomenon which does not originate from the slow information diffusion underlying previously documented lead-lag effects, weekly reversals, momentum, nonsynchronous trading, or other known factors. Our findings suggest that institutional portfolio reallocations can induce return predictability among otherwise unrelated stocks. * All three authors are from Cornell University. Gao is at the Samuel Curtis Johnson Graduate School of Management (phone: ; fax: ; pg297@cornell.edu); Moulton is at the School of Hotel Administration (phone: ; fax: ; pmoulton@cornell.edu); and Ng is at the Dyson School of Applied Economics and Management (phone: ; fax: ; dtn4@cornell.edu). We thank Warren Bailey, Paul Gao, Byoung-Hyoun Hwang, Kewei Hou, Danling Jiang, Andrew Karolyi, Eric Kelley, Charles Lee, David Musto, Lilian Ng, Gideon Saar, Zheng Sun, Huacheng Zhang, and seminar participants at the City University of Hong Kong, Cornell Finance Brown Bag Workshop, Syracuse University, University of Wisconsin at Milwaukee, the Conference on Financial Economics and Accounting, the Mid-Atlantic Research Conference in Finance, and the Midwest Finance Association Annual Meeting for helpful comments.

4 Institutional Ownership and Return Predictability Across Economically Unrelated Stocks Abstract We document strong weekly lead-lag return predictability across stocks from different industries with no customer-supplier linkages (economically unrelated stocks). Between 1980 and 2010, the industry-neutral long-short hedge portfolio earns an average of over 19 basis points per week. This return predictability arises exclusively from pairs of stocks in which there are common institutional owners. This predictability is a new phenomenon which does not originate from the slow information diffusion underlying previously documented lead-lag effects, weekly reversals, momentum, nonsynchronous trading, or other known factors. Our findings suggest that institutional portfolio reallocations can induce return predictability among otherwise unrelated stocks. Keywords: Return predictability; anomalies; institutional ownership; institutional trading JEL Classifications: G12; G14

5 The finance literature has documented that some stocks lead other stocks in returns (leadlag cross-autocorrelation) in several different contexts, including from large to small firms within the same industry, between customer-supplier linked firms and industries, from more actively to less actively traded stocks, from high to low institutional ownership stocks, and from easy-toanalyze firms to complicated firms. 1 In all of these cases, the explanation for the lead-lag effect is slow information diffusion, often among economically linked firms and industries. 2 This paper documents a new type of return predictability that is distinct from previous studies. We investigate whether common institutional ownership (that is, the same institution holding multiple stocks) is related to return predictability between the stocks of otherwise economically unrelated firms. 3 More specifically, can the historical return relations between economically unrelated stocks that have common institutional owners be used to predict the subsequent returns of a stock? Our central idea is that after observing abnormal returns for one stock in his portfolio, an institutional investor is likely to revisit his investment decisions and re-optimize his entire portfolio, which can cause him to buy or sell stocks that are unrelated to the stock whose returns motivate the portfolio changes. For example, some institutional investors have limits on how much of their portfolios can be invested in a single stock, requiring them to sell a stock whose value rises above a certain level and reallocate the funds to other stocks in their portfolios. 4 Previous theoretical (e.g., 1 These phenomena are documented in Cen, Ling, Chen, Dasgupta, and Gao (2013), Hou (2007), Lo and MacKinlay (1990), Cohen and Frazzini (2008), Huang and Kale (2013), Menzly and Ozbas (2010), Chordia and Swaminathan (2000), Badrinath, Kale, and Noe (1995), and Cohen and Lou (2012), respectively. 2 Other examples of lead-lag effects caused by slow information diffusion include studies showing return predictability from high analyst coverage to low analyst coverage stocks (Brennan, Jegadeesh, and Swaminathan, 1993), from low friction stocks to high friction stocks (Hou and Moskowitz, 2005), from illiquid large stocks to smaller stocks (Chordia, Sarkar, and Subrahmanyam, 2011), from globally accessible stocks to inaccessible stocks (Bae, Ozoguz, Tan, and Wirjanto, 2012), and from industries to the market (Hong, Torous, and Valkanov, 2007). 3 Since all firms are exposed to market shocks and macroeconomic factors such as GDP growth and inflation, they are all economically related in the broadest sense. Our definition of economically unrelated specifically focuses on links between firms cash flows that could lead to information transfers between stocks, rather than common macroeconomic fundamentals that drive all stocks returns. 4 In a related vein, Covrig, Fontaine, Jimenez-Garces, and Seasholes (2009) and Hau and Rey (2009) examine the portfolio rebalancing effect of institutions. Another example of how price changes in some stocks may 1

6 Basak and Pavlova, 2013, and Cont and Wagalath, 2014) and empirical work (e.g., Anton and Polk, 2012) finds that institutional portfolio readjustments can lead to higher return correlations among the stocks held by an institution. When capital is slow moving (Duffie 2010) or readjustments take place periodically (e.g., weekly) rather than instantaneously, the collective actions of institutional investors can give rise to price pressures and subsequent short-term return predictability (see Section 1 for discussion of related literature). 5 We exclude economically linked firms from our analysis in order to shut down the classic cash flow links between firms, allowing a clearer focus on the role of common institutional ownership. Our empirical design begins with identifying pairs of stocks that are from different industries and whose industries have no supplier-customer links ( unrelated stocks ). Following Menzly and Ozbas (2010) and Huang and Kale (2013), we identify supplier-customer links from the Bureau of Economic Analysis (BEA) Benchmark Input-Output Surveys. To verify that the stocks in each pair are not economically related, we also examine the correlation between their earnings surprises. The economically unrelated stock pairs have low earnings surprise correlations in general (average 0.017), and our results are robust to excluding stock pairs with significant earnings surprise correlations. For each pair of unrelated stocks, we examine the history of how one stock s cumulative abnormal return relates to the second, economically unrelated stock s cumulative abnormal return over a subsequent week. 6 We use cumulative abnormal returns rather than raw returns in order to remove market-wide effects. We apply the coefficient estimates from a historical regression to the induce institutional trading in other stocks is provided by Hau and Lai (2012), who find evidence consistent with mutual funds that have high exposure to financial stocks engaging in asset fire sales of non-financial stocks during the 2007 financial crisis. Similarly, Broner, Gelos, and Reinhart (2006) document how fund trading can propagate financial crises. 5 Chakrabarty, Moulton, and Trzcinka s (2015) model shows that rationally optimizing portfolio managers weigh the cost of readjustment (including price pressures) against the cost of being away from their desired portfolio allocations. Thus the mere existence of price pressures does not necessarily keep institutional investors from trading. 6 For brevity, we often refer to cumulative abnormal returns as simply returns ; all of our analyses are based on cumulative abnormal returns, as defined in Appendix A. Details of the return prediction methodology are contained in Section

7 first stock s recent performance, in order to predict the second stock s future weekly cumulative abnormal return. We then aggregate multiple return predictions (based on different economically unrelated stocks) for each stock and sort the stocks into industry-neutral portfolios based on their average predicted returns. 7 We find strong weekly return predictability from economically unrelated stocks. During the 1980 to 2010 sample period, the industry-neutral long-short hedge portfolio, which is long (short) the stocks with the highest (lowest) predicted returns, earns an average of over 19 basis points per week (with a t-statistic above five), implying an annualized average return of nearly 10%. This return predictability arises exclusively from the pairs of stocks in which there are common institutional owners. When we forecast returns using only pairs of unrelated stocks that do not share common institutional owners, we find insignificant return predictability. We examine numerous alternative explanations and find that our results are distinct from previously documented return predictability. Our results are not explained by industry or suppliercustomer linkages between firms, as we exclude all such economically related stock pairs from our analysis. Industry and sector rotation do not explain our results, since our strategy employs industry-neutral portfolios. Our findings are not explained by previously documented lead-lag relations arising from slow information diffusion, including from large to small firms, more actively traded to less actively traded stocks, high institutional ownership to low institutional ownership stocks, and high analyst coverage to low analyst coverage stocks. Our documented predictability is also distinct from well-known return anomalies including size effects, book-to-market effects, weekly and monthly return reversals, long-run reversals, price momentum, earnings momentum, liquidity effects, and trading volume effects. Our results are not due to nonsynchronous trading or seasonality. Our sample includes only stocks with share prices not less than $5 at the end of the prior quarter, and the return predictability results are qualitatively unchanged when we use only 7 Note that this is not a pairs trading strategy, although pairs of economically unrelated stocks are used to predict returns. 3

8 stocks that trade every day in the previous 12 months. Overall, we find a novel, highly robust link between common institutional ownership and return predictability for economically unrelated stocks. To investigate the mechanism through which common institutional ownership is associated with return predictability among economically unrelated stocks, we analyze changes in quarterly institutional holdings. We find that institutions accumulate more of stocks in the highest predictedreturn quintile than in the lowest predicted-return quintile, linking return predictability to institutional portfolio changes. We also explore the mechanics of the return signals arising from the economically unrelated stock pairs and find patterns consistent with cross-stock reallocations within institutions. The remainder of the paper is organized as follows. Section 1 describes related literature on institutional investors and stock returns. Section 2 presents the data and our methodology for constructing return predictions and forming portfolios. Section 3 presents the main results on return predictability among economically unrelated stocks with and without common institutional owners. Section 4 investigates other possible explanations for return predictability in economically unrelated stock pairs. Section 5 examines how the predictability relates to changes in institutional holdings. Section 6 discusses additional analyses and robustness checks, and Section 7 concludes. Appendix A contains variable definitions. Appendix B provides a detailed example of how return predictions are determined using a specific pair of unrelated stocks. 1. Relation to literature on institutional investors and stock returns Our paper builds on and contributes to recent literature on how institutional ownership may affect stock return variance or correlations. Greenwood and Thesmar (2011) show that fragile stocks (i.e., stocks with high percentages held by a few institutions) have high volatility. Anton and Polk (2012) show that the degree of common institutional ownership forecasts cross-sectional 4

9 variation in return correlation, and Bartram, Griffin, Lim, and Ng (2013) show that foreign ownership linkage is an important driver of the covariation of returns for stocks in different countries. These studies suggest that simply by investing in multiple stocks, institutional investors may affect the contemporaneous return correlations between stocks. 8 In addition, capital can be slow moving as Duffie (2010) suggests. Institutions may re-adjust their portfolios periodically (for example, weekly) rather than instantaneously. In this case, common institutional ownership could affect not only contemporaneous return correlations but also cross-autocorrelations of stocks they hold. Our study focuses on the unexamined question of whether common institutional investment is associated with lead-lag return predictability across different stocks. By examining the return predictability of economically unrelated stocks owned by the same institution, our study focuses on common institutional ownership in the absence of information links between firms. Our paper is also related to, but distinct from, recent papers that emphasize fund flows as a mechanism that creates price pressure. Coval and Stafford (2007) find that funds experiencing large outflows create price pressures on stocks held in common by distressed funds, while Jotikasthira, Lundblad, and Ramadorai (2012) document fire sale effects in emerging markets. Frazzini and Lamont (2008) find that mutual fund flows negatively predict future long-term returns, while Lou (2012) finds that the mutual fund flow-driven return effect can partially explain stock price momentum. While these papers focus on how capital inflows and outflows induce trading that in turn affects stock prices, our paper investigates a different mechanism: the reallocation of institutional investors capital from some stocks to others, which occurs whenever institutional investors adjust their portfolios, not only when large inflows or outflows occur. This work contributes to our understanding of how institutional trading may affect returns. Previous empirical papers have documented that mutual fund herding may move the price of small stocks in subsequent quarters (Wermers, 1999; Sias, 2004). There are generally three explanations 8 Similarly, the accounting literature documents evidence of institutional ownership affecting stock returns around earnings announcements (e.g., Potter, 1992; Bartov, Radhakrishnan, and Krinsky, 2000). 5

10 offered for why institutional trading may affect subsequent stock returns (Sias, Starks, and Titman, 2006). One is that institutions uncover private information about individual stocks and reveal it through their trading, leading to permanent price effects (e.g., Easley and O Hara, 1987; Kyle, 1985; Boehmer and Kelley, 2009). A second explanation for a permanent price effect from institutional trades is that investors view stocks as imperfect substitutes and their long-term supply and demand curves are not perfectly elastic. Thus the non-institutional traders who are on the other side of aggregate institutional trades demand lower (higher) prices to buy (sell) stocks (e.g., Shleifer, 1986; Bagwell, 1991; Lynch and Mendenhall, 1997; Greenwood, 2005). The third explanation implies a temporary price effect from institutional trading. Institutional trading may affect stock prices if it pushes liquidity providers away from their preferred inventory position (e.g., Stoll, 1978; Grossman and Miller, 1988) or if there is slow movement of investment capital to trading opportunities (Duffie, 2010). We find that the return predictability from unrelated stocks is a temporary price effect, yielding the highest return in the first week after portfolio formation and then reversing in the following weeks. This pattern suggests that the return predictability arises primarily because aggregate trading from institutional portfolio adjustments results in temporary price pressures, rather than because institutions are trading on superior information or long-term supply and demand curves for non-institutional traders are elastic. 2. Data and methodology Our analysis uses stock return data from the Center for Research in Security Prices (CRSP), earnings announcement and accounting data from Compustat, analyst forecast data from the Institutional Brokers Estimate System (I/B/E/S) from Thomson Reuters, 13F institutional holdings data from Thomson Reuters, and information on customer-supplier industry links from the Bureau 6

11 of Economic Analysis (BEA) Benchmark Input-Output Surveys. 9 Our sample period is January 1980 to December 2010; we start our sample in 1980 because that is when the institutional holdings data begin. We begin with the universe of all common stocks (CRSP share codes 10 and 11) listed on NYSE, AMEX, and NASDAQ, and apply the following screens to create our sample of weekly observations: the share price at the end of the previous quarter must be greater than or equal to $5; the firm must be present in Compustat data for at least the prior two years; and the most recent earnings announcement must be regular and on time (i.e., the firm makes four quarterly earnings announcements each year and has earnings announced during the three-month period after the end of each fiscal quarter). In Appendix A we provide a description of all variables used in our empirical analyses. 2.1 Economically unrelated stock pairs Pairs of economically unrelated stocks are the focus of this study. Each stock whose return we are interested in predicting ( target stock ) is matched with multiple economically unrelated stocks ( unrelated stocks ) as follows. For each target stock in our sample each week, we identify all other stocks that do not have the same Fama-French 30 industry classification. We then determine the industry code for each stock and retain only those stocks (unrelated stocks) that are from industries that show zero dollar value of inputs/outputs between them and the industry of the target stock in the most recent BEA survey. 10 We thus make sure that we are considering only stock pairs that have no industry or cash flow links between them. We recognize that despite these precautions, there still could be some more subtle economic relations between any two firms, so in 9 The BEA data are publicly available at In using the BEA data we follow Menzly and Ozbas (2010), who point out that the BEA surveys provide a more complete picture in identifying economically related stocks than the Compustat customer information database used in Menzly and Ozbas (2004) and Cohen and Frazzini (2008). 10 We use BEA s standard make and use tables at the detailed level, which identify 484, 496, 537, 542, 498, 498, 511, and 537 industries in the surveys from 1967, 1972, 1977, 1982, 1987, 1992, 1997, and 2002, respectively. The BEA survey uses Standard Industry Codes (SIC) prior to 1997 and North American Industry Classification System (NAICS) codes from 1997 on. We merge SIC and NAICS codes as in Menzly and Ozbas (2010). 7

12 our robustness checks we exclude any stock pairs that have significant unexpected earnings correlations over our sample period (see Section 6.3). 2.2 Institutional ownership We count the number of common institutional owners and the number of significant common institutional owners for each pair of economically unrelated stocks using the quarterly 13F institutional holdings data. Common institutional owners are defined as the same institutional investor holding positions in two stocks as of the prior quarter-end. Because we expect that any return predictability connected to institutional trading should be stronger when institutions have larger common holdings, we further categorize common institutional ownership as significant if an institution holds more of each stock than the median institutional holder of that stock. For example, if the median institutional holding in stock A is 0.4% of shares outstanding and the median institutional holding in stock B is 0.1%, we define an institution that holds more than 0.4% of stock A and more than 0.1% of stock B as a significant common owner. For our main analyses, we identify a stock pair as having common institutional ownership if it has common institutional owners on all of the prior 20 quarter-ends, while a stock pair would have no common institutional ownership if it has no common institutional investor in any of the prior 20 quarter-ends. 11 Similarly, a pair of stocks would have no significant common institutional owners if there is no significant common institutional investor in any of the past 20 quarter-ends. 2.3 Sample descriptive statistics Table 1 presents descriptive statistics for the stocks in our sample. Our sample comprises 13,109 stocks. Panel A of Table 1 shows that institutions hold 41.2% of a firm s outstanding stock 11 Results are robust to determining common ownership based on institutional holdings as of only the prior quarter-end, as in Anton and Polk (2012), rather than over the prior 20 quarters. We focus on the prior 20 quarters for our main analyses because our predicted returns are based on a trailing five-year regression analysis, as described in Section 2.4 below. 8

13 on average, and the average firm has 93 institutional investors. Panel B provides some basic statistics about the pairs of economically unrelated stocks. On average there are 215 unrelatedstock pairs for each target stock. 12 Of the average 215 economically unrelated pairs of stocks, 188 have significant common institutional investors and 206 have common institutional investors. We note the small number of economically unrelated pairs that have no significant common institutional owners (an average of 20) or no common institutional owners (an average of 10); in our robustness checks we verify that the small number of pairs in these categories does not drive our results. On average there are 10 significant common institutional investors per stock pair and 27 common institutional investors per stock pair. The prevalence and variation of institutional ownership make this a promising sample in which to examine the link between institutional ownership and return predictability. [Table 1 here] 2.4 Return prediction and portfolio formation methodology The underlying mechanism we envision is that information about a specific economically unrelated stock affects investment decisions in the target stock because the portfolio manager reoptimizes his entire portfolio, trading many stocks (including the target stock), not just the one whose price has changed. To determine which stock in an unrelated stock pair is the target stock (whose return is being predicted) and which is the unrelated stock (whose return is used to predict the target s return), we determine which stock has had a more recent earnings announcement. 13 The stock with the most recent earnings announcement date is designated the unrelated stock and used to predict the return of the target stock. 12 In our robustness checks we test whether predictability is significantly affected by the number of economically unrelated pairs available. 13 Other information events, such as dividend payout announcements, merger and acquisition announcements, and idiosyncratic firm news, do not occur with enough regularity to facilitate broad analysis. Note that in this exercise we are not exploiting the information transfers among firms within the same industry, because we explicitly exclude stock pairs from the same or related industries. 9

14 To illustrate our return prediction methodology, consider one target stock and one unrelated stock in a particular week. We count the number of full weeks since the unrelated stock s last earnings announcement and the number of full weeks since the target stock s last earnings announcement; see Figure 1. Figure 1: Timing for return prediction methodology Weeks since Last EA of Unrelated Target CAR to Predict Last EA of Target Last EA of Unrelated t=0 t=1 Weeks since Last EA of Target We then search the previous five years to find occasions when the unrelated stock was exactly the same number of weeks past its most recent earnings announcement and the target stock s last earnings announcement was at least one week prior to the unrelated stock s. 14 For each occasion in the last five years, we calculate the unrelated stock s average cumulative abnormal return (CAR) over its post-earnings-announcement weeks until the week of interest. 15 We also calculate the CAR for the target stock over the subsequent week. We regress the target stock s subsequent-week CAR on the average CAR of the unrelated stock over the previous weeks since the unrelated stock s earnings announcement. We use data 14 In using five years of earnings history we follow the literature on earnings releases and anomalies (e.g., Foster, Olsen, and Shevlin, 1984). The target stock is not allowed to have more than 12 post-earningsannouncement weeks because firms generally announce their quarterly earnings every three months. 15 We require at least three years of earnings announcement date history and at least 10 valid occasions to estimate the regression for a stock pair. Since firms generally time their earnings announcements similarly across quarters and years, we typically find an adequate number of valid occasions. 10

15 from all eligible historical periods to run the regression and obtain the coefficients. We apply the coefficients from the historical regression to the average CAR for the unrelated stock over its current post-earnings-announcement weeks to predict the target stock s return for the next week. Appendix B contains a detailed example of the prediction methodology for one pair of economically unrelated stocks. For each target stock, we calculate the predicted CAR for the next week based on each of its unrelated stocks. We calculate the target stock s average predicted CAR for the next week as the mean of the predictions from all of its unrelated stocks. 16 Note that since the predicted return and the stock returns used to predict it are all cumulative abnormal returns, market-wide effects are already removed. Finally, we form industry-neutral quintile portfolios based on the target stocks predicted returns, to prevent our results from being driven by industry rotation. 17 We repeat this procedure for all stocks each week to form weekly quintiles based on predicted weekly returns. Table 2 presents descriptive statistics for the quintile portfolios, which form the basis for our tests. We calculate simple averages of firm-level characteristics for stocks within each quintile and then report time-series averages of each quintile s characteristics from January 1980 to December Panel A shows that there are variations across the predicted-return portfolios in terms of the component stocks size, book-to-market, lagged returns, volatility, liquidity, and trading volume. For example, the predicted-car-sorted portfolios are monotonically decreasing in book-to-market and increasing in price momentum over the previous year (Return month t-12 to 16 We also consider the weighted mean of the predictions from unrelated stocks where weights are based on the precisions of predicted values. The results are qualitatively similar to those based on the simple average. 17 To form industry-neutral quintile portfolios, we identify all of the target stocks by their Fama-French 30 industries, and within each industry we sort the target stocks into five groups (each containing 20% of the stocks in that industry): Group 1 contains the target stocks with the lowest predicted returns for the following week, and Group 5 contains the target stocks with the highest predicted returns for the following week. We then form industry-neutral portfolios by combining the target stocks in Group 1 from all 30 of the Fama- French industries into a single Quintile 1 portfolio, and similarly with the remaining four groups to form the five industry-neutral predicted-return portfolios. 11

16 t-2). Therefore in addition to investigating the potential relation of return predictability to common institutional ownership, we examine other previously documented explanations for return predictability, such as size, book-to-market, momentum, reversals, and liquidity, in Section 4. Panel B shows that return autocorrelations are generally small in each of the quintile portfolios. [Table 2 here] 3. Return predictability and common institutional ownership Table 3 presents our core results on return predictability from economically unrelated stock pairs, using all economically unrelated stock pairs to predict returns. We report the value-weighted weekly excess return (ER) and value-weighted alphas from Capital Asset Pricing Model (CAPM), Fama-French three-factor (FF3), and Fama-French-Carhart four-factor (FFC4) regressions for each portfolio (Fama and French, 1993; Carhart, 1997). 18 The excess returns and alphas are reported in percent; for example, the excess return of for Quintile 1 represents 2.0 basis points per week. Quintile 1 (5) is an industry-neutral portfolio containing stocks with the lowest (highest) predicted returns, and the bottom row tests the return difference between Quintile 5 and Quintile 1 (Q5-Q1 spread), the classic long-short portfolio. We find strong return predictability in this long-short portfolio, with weekly excess return and Fama-French-Carhart alpha both over 19 basis points and t-statistics of 5.3 and 5.7, respectively All results are qualitatively similar when returns are equal-weighted rather than value-weighted. We use value-weights rather than equal-weights in the calculation of daily portfolio returns for the following three reasons: (1) equal-weighting of daily returns leads to portfolio returns that may be overstated because of the so-called bid-ask bounce effect (see Blume and Stambaugh, 1983; and Canina, Michaely, Thaler, and Womack, 1998); (2) equal-weighting of daily returns essentially assumes daily rebalancing of portfolios, which could further overstate the economic magnitude of the returns: and, (3) value-weighting of daily returns better captures the economic significance of the covariance implied returns because equal-weighting of returns over-represents smaller firms. Value-weighting may bias against finding any evidence of abnormal returns, since stocks with larger market capitalization are more likely to be informationally efficient -- including efficiency in the incorporation of information from the early announcers. 19 Factor loadings from the four-factor Fama-French-Carhart regressions are reported in the Internet Appendix. 12

17 [Table 3 here] Figure 2 graphs annual long-short portfolio returns and Sharpe ratios over the 31-year period. The long-short hedge portfolio annual return is calculated as the average weekly excess return times the number of weeks in the year. The Sharpe ratio divides the annual excess return by the annualized standard deviation of weekly returns. The annual return is positive in all but three years of the sample period and notably remains positive even during the financial crisis in [Figure 2 here] We next calculate predicted returns using subsets of the economically unrelated stock pairs. In Panel A of Table 4, we calculate predicted returns for each stock first using only stock pairs that have significant common institutional owners, and then using only stock pairs that have no significant common institutional owners. Panel A compares the weekly return performance of each set of predicted return quintile portfolios. Portfolios formed based on predicted returns from stocks with significant common institutional owners (the first four columns) show strong predictability: The excess return difference in the long-short portfolio is 19.8 basis points per week with a t- statistic of 4.8. In contrast, portfolios formed using only predicted returns from stocks with no significant common institutional owners (the middle four columns) show insignificant return spreads in the long-short portfolio: spreads of less than five basis points per week with t-statistics just over one. The final four columns show that the difference between the long-short portfolio returns using stock pairs with versus without significant common institutional owners is also significant. The excess return difference is estimated at 15.5 basis points, risk-adjusted alphas are of similar magnitude, and all are significant. Panel B shows that basing return predictions on stock pairs with common institutional owners yields similar results. The difference between long-short portfolio returns based on stock pairs with versus without common institutional owners is 16.5 basis points in excess return with a t-statistic of 2.2. All of these results support our conjecture that 13

18 return predictability among economically unrelated stocks is related to common institutional ownership. [Table 4 here] We extend our analysis of one-week return predictability with versus without significant common institutional investors by calculating the weekly Fama-French-Carhart 4-factor alpha for each portfolio and the long-short portfolio strategy from one week to twelve weeks after portfolio formation. Figure 3 shows that the first-week alpha is much higher for the long-short strategy based on stock pairs with significant common institutional investors, as in Table 4, Panel A. The abnormal returns dissipate quickly and are reversed in the following weeks, leading to cumulative average weekly returns near zero for the longer holding periods, consistent with institutional investors trading patterns creating temporary price pressures (e.g., Stoll, 1978; Grossman and Miller, 1988; Duffie, 2010). The timing of the reversals over the following several weeks is consistent with that of other reversals documented in the literature (e.g., Cohen and Lou, 2012; Coval and Stafford, 2010). [Figure 3 here] 4. Alternative explanations for return predictability In this section we examine whether our findings on the link between common institutional ownership and return predictability could simply be a manifestation of other factors that are already known to be related to return predictability. Table 5 examines lead-lag effects. Return predictability is known to be related to the relative size of firms, with large-firm returns leading small-firm returns (e.g., Lo and MacKinlay, 1990; and Hou, 2007). Panel A shows that our predictability results are significant when predicted returns are calculated based on stock pairs in which the target firm is from a larger or equal size 14

19 decile than the economically unrelated stock used to predict its return (left panel) and when the target firm is smaller (right panel). 20 Cohen and Lou (2012) find that the returns of stand-alone firms, which operate in only one industry, can be used to predict the returns of conglomerates, which are involved in multiple industries but are assigned a single SIC code that reflects the firm s main industry segment. To verify that our results are not driven by the presence of conglomerates, we run our analysis using stock pairs in which all of the stocks are stand-alone firms and, separately, all of the stocks are conglomerates. 21 Panel B shows that our results are robust to excluding conglomerates. Chordia and Swaminathan (2000) find a lead-lag effect from more actively traded stocks to less actively traded stocks. Panel C shows that our predictability results are significant when predicted returns are calculated separately based on stock pairs in which the target firm is from a larger or equal NYSE/AMEX volume decile than the economically unrelated stock (left panel) and when the target firm is from a smaller volume decile (right panel). Panel D presents analogous results using pairs of NASDAQ stocks. Badrinath, Kale, and Noe (1995) find that stocks with high institutional ownership lead the returns of stocks with low institutional ownership. Panel E shows that our predictability results are significant when predicted returns are calculated separately based on stock pairs in which the target firm has higher or equal decile institutional ownership than the economically unrelated stock (left panel) and when the target firm has lower decile institutional ownership (right panel). Brennan, Jegadeesh, and Swaminathan (1993) find return predictability from stocks with high analyst coverage to stocks with low analyst coverage. Panel F shows that our results are robust to using stock pairs in which the target firm has higher or equal (left panel) or lower (right panel) 20 Hou (2007) shows that the lead-lag effect from large firms to small firms is mainly driven by intra-industry effects; since we exclude firm pairs within the same industry, our findings are not a contradiction of his. 21 We thank Dong Lou for sharing his list of conglomerates and stand-alone firms. 15

20 analyst coverage than the unrelated stock. In short, none of the previously documented lead-lag relationships explain our results. [Table 5 here] We next conduct double portfolio sorts in Table 6 to examine whether other documented return anomalies can explain our results. Previous literature has documented return anomalies due to size and book-to-market (Fama and French, 1992), past 1-week returns (Lehmann, 1990), past 1-month returns (Jegadeesh, 1990), past 12-month returns up until the prior month (Jegadeesh and Titman, 1993), earnings surprise (Chan, Jegadeesh, and Lakonishok, 1996; Sadka, 2006), illiquidity (Amihud, 2002), and trading volume (Chordia and Swaminathan, 2000). A natural question is whether our results could be due to one of these well-documented effects rather than institutional ownership per se (for example, we know from Table 2 that our predicted-car-sorted portfolios are monotonically decreasing in book-to-market and increasing in prior-year price momentum). Thus our interest in these double sorts is whether we find predictability (significant Quintile 5 minus Quintile 1 differences) within secondary sorts on each of these stock characteristics. Table 6 reports double sorting results in which we conduct independent, industry-neutral sorts on predicted CARs and firm size, book-to-market equity, weekly return reversals, monthly return reversals, momentum, long-run return reversals, earnings momentum, liquidity, and NYSE/AMEX trading volume and turnover. We find that our return predictability is robust to all of these secondary portfolio sorts. In particular, our results are not driven by weekly or monthly return reversal or momentum effects. For brevity, we report only the top and bottom predictability quintiles and their differences in Table 6; full results are reported in the internet appendix. [Table 6 here] Next we conduct Fama-MacBeth (1973) cross-sectional regressions to test whether the return predictability arising from economically unrelated stocks remains significant in a multivariate setting that includes explanatory variables previously linked to return predictability. 16

21 Table 7 presents the results of the time-series average of Fama-MacBeth regression coefficients (and t-statistics) when we regress stocks weekly excess returns on the previously predicted CARs. In particular, we use each stock s predicted-return quintile number (5=highest predicted return, 1=lowest) as the first explanatory variable, and we include other known explanatory factors as control variables in alternate specifications. In specifications (1) and (2), the predicted CARs are based on all economically unrelated stock pairs. The coefficient on the predicted CAR is positive and highly significant, showing that the predictability documented in this paper is not subsumed by return reversals, price momentum, earnings momentum, or other firm characteristics including market capitalization, book-to-market equity, operating accruals, net stock issuance, idiosyncratic volatility, or Amihud illiquidity. In specifications (3) and (4), the predicted CARs are based on only economically unrelated stock pairs with significant common institutional owners. The coefficients on predicted CAR remain positive and highly significant. In contrast, the coefficients on predicted CAR are insignificant in specifications (5) and (6), where the predicted CARs are based on pairs with no significant common institutional owners. Taken together, the results in Tables 5, 6, and 7 confirm that the return predictability is driven by stock pairs with significant common institutional owners and is not subsumed by other documented sources of predictability. [Table 7 here] 5. Institutional portfolio changes and return predictability In this section, we analyze institutional portfolio changes to see whether they are consistent with our notion of how common institutional ownership is related to return predictability between economically unrelated stocks. We examine the changes in quarterly institutional holdings of stocks in the high versus low predicted return quintiles. We posit that institutions increase their holdings more in stocks that rank in the highest predicted return quintile (Quintile 5) than stocks in the lowest 17

22 predicted return quintile (Quintile 1). Quintile portfolios are formed weekly, but institutional holdings are reported only quarterly, so we focus on stocks that are consistently ranked in the same quintile throughout the calendar quarter. Table 8 presents the change in percentage institutional ownership for stocks consistently ranked in each predicted return quintile during the same calendar quarter. [Table 8 here] Panel A of Table 8 presents the results for stocks that are in the same quintile portfolio for at least 75% of the weeks in the quarter. Panel B takes an alternative definition of consistently, based on stocks that are in the same quintile portfolio for at least 50% of the weeks in the quarter. Overall, the results show that institutional investor ownership increases more for stocks with the highest predicted returns (Quintile 5) than for those with the lowest predicted returns (Quintile 1). For example, Panel A shows that the average change in percentage institutional ownership is more than one percentage point greater for Quintile 5 than for Quintile 1 stocks, and the difference is significant. This evidence is consistent with the notion that institutional trading activity induces the return predictability among economically unrelated stocks. 6. Additional analyses and robustness checks 6.1 Positive versus negative predicted returns The intuition behind our empirical set-up is that after observing the return on one stock he owns, a portfolio manager decides to buy or sell a different, economically unrelated stock. If he wants to buy, he could either buy more of a stock he already owns or buy another stock, but if he wants to sell, his choices are likely limited to stocks he already owns. Thus stocks may experience more selling rather than buying pressures from owners who reallocate from their other portfolio holdings. Based on this intuition, we might expect the return predictability to be stronger for stocks 18

23 when the signals from the economically unrelated stocks suggest selling (i.e., a negative predicted return). In Table 9, we predict returns separately for stocks using only the signals from economically unrelated stocks with significant common institutional investors that predict a negative return versus only signals that predict a positive return. As we are using only a subset of the available signals (positive or negative) for each stock, we expect the resulting predicted return for each stock to be noisier and the Quintile 5 minus Quintile 1 spread within each subset to be less significant. [Table 9 here] Panel A shows some predictability when we use only the negative signals, but we find no predictability when we use only the positive signals. Using only pairs predicting negative returns, the Quintile 5 minus Quintile 1 spread after controlling for Fama-French and Fama-French-Carhart factors is significant. In contrast, using only pairs predicting positive returns, none of the Quintile 5 minus Quintile 1 differences are significant. These subset results are consistent with our expectation that return predictability is stronger for stocks when the signals from the economically unrelated stocks suggest selling (i.e., a negative predicted return). In Panel B we investigate the negative-predicted-return signals in more detail, bearing in mind that the more finely we separate the subsets the harder it is to see any predictability within the subset. Negative-predicted-return signals can arise in two ways. First is an unrelated-stock-loss channel. After one stock s price declines, institutional investors may sell another stock in order to reduce their equity exposure (similar to Kodres and Pritsker, 2002) or to meet liquidity demands (Coval and Stafford, 2007). In our empirical set-up, such cases arise when two stocks have a positive historical correlation and the unrelated stock has a negative recent return. Second is a return-chasing channel. After one stock s price rises, in order to increase their investment in that stock institutional investors may sell another stock (Bohn and Tesar, 1996). In our setting, this happens when two stocks have a negative historical correlation and the unrelated stock has a 19

24 positive recent return. Panel B presents the results separately for the unrelated-stock-loss and return-chasing channels. We find some support for the unrelated-stock loss channel (Fama-French and Fama-French-Carhart Q5-Q1 spreads are significant) and less support for the return-chasing channel (only Fama-French Q5-Q1 spread is significant). 6.2 Seasonality Previous literature suggests that institutional investors are more concerned about readjusting their portfolios at certain times of the year, such as at quarter-ends and month-ends (e.g., Lakonishok, Shleifer, Thaler, and Vishny, 1991; Moulton, 2005). In our context, we thus expect to see stronger predictability arising at quarter- and month-ends, and a natural question is whether all of the predictability is driven by the last week of the month or quarter. In Panel A of Table 10 we separate out the last week in each calendar quarter, and in Panel B we separate out the last week in each month. We find larger Quintile 5 minus Quintile 1 return differences in the end-of-quarter and end-of-month weeks than other weeks, although the non-end-of-quarter and non-end-of-month differences remain significant. The higher predictability in quarter-end and month-end weeks is consistent with increased portfolio adjustments at those calendar intervals. [Table 10 here] 6.3 Narrower definition of economically unrelated stocks Even though the stock pairs we identify have no direct cash flow links (since they are from different industries that have zero dollar value in the standard BEA make-use tables at the detailed level), it is possible that they could have some more subtle economic links that our methodology does not capture. To account for this possibility, we examine the correlations between unexpected earnings for each pair of economically unrelated stocks over our entire sample period. As Panel A of Table 11 shows, the average correlation is only About 9.5% of the stock pairs in our 20